3/30/12

Four-dimensional
computed
tomography
(4DCT)
George Starkschall, PhD
Peter Balter, PhD
MD Anderson Cancer Center

Disclosure  
Some of the authors’ work cited in
this presentation was supported
by a Sponsored Research
Agreement with Philips Medical
Systems

Questions?
•  Please text questions to (713) 906-7259

1
 3/30/12

Educational Objectives  
At the conclusion of this presentation, the
participant will be able to answer the following
questions:

1.  What do we mean by 4DCT?

2.  Why would we need 4DCT?
3.  How do we obtain 4DCT images?
4.  How do we use 4DCT images?
5.  How can we improve the quality of 4DCT
images?

What do we mean by 4DCT?

•  4DCT is the acquisition of dynamic

anatomic information via CT in which
we use periodic motion to replace the
time dimension with the phase
dimension.

Time scales
•  Typical respiratory cycle - ~4-5 sec
•  Typical CT acquisition time - ~5 sec
(multislice helical CT)
•  How can we capture sub-respiratory
cycle information given that the typical
CT acquisition time is approximately
equal to the respiratory cycle time?

2
 3/30/12

Resolution of problem
•  Take advantage of periodic nature of
respiratory cycle
•  Acquire small amount of information during
one cycle, more information the next cycle,
more information the next cycle …
•  Combine information from multiple
respiratory cycles
•  The 4th dimension becomes phase, rather
than time

Why would we need 4DCT?

•  20th century radiation treatment planning

–  Population-based radiation therapy
–  Account for respiratory motion by adding
uniform, isotropic internal margin (IM) to
clinical target volume (CTV) to generate
internal target volume (ITV)

Why would we need 4DCT?

•  Assumptions:
–  Lung tumors move the same amount,
irrespective of patient, location, size, etc.
–  Lung tumor margins expand and contract
isotropically

3
 3/30/12

Why would we need 4DCT?

•  Assumptions:
–  Lung tumors move the same amount,
irrespective of patient, location, size, etc.
–  Lung tumor margins expand and contract
isotropically

•  Neither of these assumptions is true!

Red Journal 53:822-834 (2002)

How much do thoracic tumors

move ?
Tumor Motion Histogram

20
18

16
14
# of tumors

12
10
8
6
4

2
0
0

0.0 - 0.2

0.2 - 0.4

0.4 - 0.6

0.6 - 0.8

0.8 - 1.0

1.0 - 1.2

1.2 - 1.4

1.4 - 1.6

1.6 - 1.8

1.8 - 2.0

Tumor Motion (cm)

Red Journal 68: 531-540 (2007)

4
 3/30/12

CT scan times

•  Prior to the development of helical CT,

scan times were in excess of 30 sec,
encompassing several respiratory
cycles
•  Single slice randomly sampled
respiratory cycle
–  Resulting in sloppy images

Why would we need 4DCT?

•  21st century radiation treatment planning

–  Personalized radiation therapy
–  Account for respiratory motion by adding
patient-specific internal margin (IM) to
clinical target volume (CTV) to generate
internal target volume (ITV)

How do we obtain 4DCT images?

•  Enabling technologies
–  High-speed CT – multislice helical
–  Respiratory monitoring
•  Two approaches
–  Image binning [Pan, et al, Med Phys
31:334-340 (2004)]
–  Projection binning [Keall, et al, PMB
49:2053-2067 (2004)]

5
 3/30/12

General approach to 4DCT image

acquisition

•  Acquire image data continuously during

respiration
•  Reconstruct the image data at specific
phases in the respiratory cycle for each
patient location.
•  Combine image data at same phase from
several respiratory cycles.
•  Result: A series of 3DCT images, each
representing a different phase in the
respiratory cycle.

Two approaches to respiratory

monitoring

Measurement of Measurement of
abdominal height abdominal circumference

The Varian RPM® system

IR Camera

IR Reflectors RPM software tracks the markers

Can be used to monitor patient
breathing and for patient feedback

6
 3/30/12

Image binning

•  Operate CT scanner in cine mode

•  Monitor respiratory cycle
•  Acquire multiple images with table in
fixed position
–  Time period > 1 respiratory cycle + 2
gantry rotations
–  Ensures adequate sampling
•  Index table
•  Repeat image acquisition

4D-CT Data Acquisition

X-ray tube

CT det.

Animation by:
Tinsu Pan, Ph.D

4D-CT Data Acquisition

X-ray tube

CT det.

Animation by:
Tinsu Pan, Ph.D

7
 3/30/12

4D-CT Data Acquisition

X-ray tube

CT det.

Animation by:
Tinsu Pan, Ph.D

Cine based 4DCT

Infrared
Infrared Camera
Reflector

Image binning

•  Obtain large number (1000-3000) of

images, each figure associated with an
acquisition time

Pan, et al, 2004

8
 3/30/12

Image binning

•  Determine acquisition times associated

with corresponding phases in
respiratory cycles
–  Typically 10 phases (0%, 10%, 20%, …)
•  Bin images with acquisition times
closest to phase acquisition time

Projection binning

•  Operate CT scanner in helical mode at

very low pitch
–  Table translation during one respiratory
cycle < detector width
•  Monitor respiratory cycle

Projection binning

•  Obtain large number of projections,

each projection associated with an
acquisition time

9
 3/30/12

Projection binning

•  Determine acquisition times associated

with corresponding phases in
respiratory cycles
–  Typically 10 phases (0%, 10%, 20%, …)
•  Bin projections with acquisition times
closest to phase acquisition time
•  Reconstruct images based on binned
projections

4DCT
4D CTbased
basedon
onaaSpiral
SpiralCT
CTAcquisition
Acquisition
10
20
30
40
50
60
70
80
0% %Phase
Phase––Sampled
Sampledby bydetector
detector19
2
3
4
5
6
7
8

An oversimplification for demonstration of data sufficiency requirements

Result of image acquisition

•  In both cases, the result is a set of

several (typically 10) three-dimensional
CT data sets, each data set
corresponding to a different phase of
the respiratory cycle.

10
 3/30/12

Helical 4DCT

What about patient dose?

•  Li et al [Med Phys 32:3650-3660 (2005)]

–  4DCT of chest gives dose of ~200 mSv
•  Keall et al [Phys Med Biol 49:2053-2067
(2004)]
–  250-400 mSv
•  Mayo et al [Radiology 228:15-21 (2003)]
–  conventional CT dose of ~7 mSv

How do we used 4DCT images?

•  Recall ICRU definitions

–  Gross tumor volume (GTV)
–  Clinical target volume (CTV)
–  Internal target volume (ITV)
–  Planning target volume (PTV)

ICRU 50,62

11
 3/30/12

Gross Tumor Volume (GTV)

•  “Gross demonstrable
extent and location of
the malignant
growth”
o  Primary tumor
o  Involved nodes
o  Metastatic disease
•  Acquire information
from imaging study

Clinical Target Volume (CTV)

•  “Tissue volume that

contains a
demonstrable GTV
and/or subclinical
malignant disease
that must be
eliminated”
•  Expand GTV based
on knowledge of
disease spread

Internal Target Volume (ITV)

•  CTV + internal margin (IM) to

compensate for all movements
•  Respiration
•  Bladder and rectum fillings
•  Swallowing
•  Cardiac motion
•  Bowel motion
•  Expand CTV based on explicit
knowledge of internal motion

12
 3/30/12

Planning Target Volume (PTV)

•  “A geometrical concept
used for treatment
planning, defined to
select appropriate beam
sizes and beam
arrangements to ensure
that the prescribed dose
is actually delivered to
the CTV”

Planning Target Volume (PTV)

•  PTV is ITV + setup margin (SM) to account

for setup uncertainties
-  Expansion of ITV based on knowledge of setup
uncertainties
-  Immobilization devices
-  On-line imaging

Target volumes

•  GTV and CTV are oncological entities

–  Independent of modality of treatment
–  Define prior to planning of treatment
•  ITV and PTV are treatment planning entities
–  Account for characteristics of the patient and
treatment methodology
–  Relevant for targeting purposes
•  The GTV and CTV belong to the radiation
oncologist; the ITV and PTV belong to the
medical physicist.

13
 3/30/12

Explicitly accounting for motion

•  The right (i.e., ICRU) way:

–  Delineate the GTV on all data sets that
comprise the 4D data set
–  Can use deformable propagation to make
life easier

Model-based deformation

•  Represent
anatomic
structure as
triangulated
mesh
•  Deform mesh
to image

Model-based deformation

•  Originally designed as method of

automatic segmentation of CT images –
deformation of library-based anatomic
structure
–  Pekar V, et al, “Automated model-based organ
delineation for radiation therapy planning in the
prostate region,” Int J Radiat Oncol Biol Phys
60:973-980 (2004)

14
 3/30/12

Model-based deformation

•  Recognized application in 4D CT
segmentation – use reference phase
contours as “library structure” and
deform to remaining phases
–  Ragan D, et al, “Semiautomated four-dimensional
computed tomography segmentation using
deformable models,” Med Phys 32:2254-2261
(2005)

Explicitly accounting for motion

•  On each phase, expand GTV to

generate CTV
–  8 mm for adenocarcinoma
–  6 mm for squamous cell carcinoma

–  Giraud P, et al, “Evaluation of microscopic tumor

extension in non-small-cell lulng cancer for three-
dimensional conformal radiotherapy planning,”
Red Journal 48:1015-1024 (2000)

Explicitly accounting for motion

•  On each phase, edit CTV to avoid

regions where there is no microscopic
extension
–  Chest wall
–  Across lobar boundaries

15
 3/30/12

Explicitly accounting for motion

•  Combine CTV from each phase to

generate ITV
•  Expand ITV to account for setup
uncertainties to generate PTV
–  MDACC thoracic service conventions
•  1 cm for conventional setup
•  0.5 cm for kV image-guided setup (based on alignment of
vertebral bodies)
•  0.3 cm for CT image-guided setup (based on alignment
of GTV)

•  Use PTV to generate treatment portals

Explicitly accounting for motion

•  Calculate dose on each phase

•  Combine CTV from each phase to
generate ITV
•  Expand ITV to account for setup
uncertainties to generate PTV
•  Use PTV to generate treatment plan

Explicitly accounting for motion

•  Create 4D plan
–  Copy original trial onto all data sets in plan
•  Compute 4D dose
–  Compute dose distribution on all data sets
–  Very time consuming and resource consuming
•  Accumulate 4D dose
–  Deform phases of 4D dose matrix to reference
phase

16
 3/30/12

Explicitly accounting for motion

•  The practical way:

–  Make use of observation that lung
parenchyma is of lower density (~0.3) than
tumor (~1)
–  Maximum intensity projection (MIP)
•  In each voxel, use maximum CT value over all
phases of 4D data set

Abstractions of 4D Data: The average dataset and

the maximum intensity projection (MIP) dataset.

Moving tumor Average showing time MIP showing all voxels

averaging of the moving occupied by the tumor over
tumor the respiratory cycle.

Explicitly accounting for motion

•  Use MIP to generate “internal gross

tumor volume” (IGTV)
–  Note that this is not an ICRU term
•  Expand IGTV to account for
microscopic disease to generate ITV
•  Expand ITV by setup margin to
generate PTV
•  Design treatment portals based on PTV

17
 3/30/12

Explicitly accounting for motion

•  Perform dose calculations on AVG data

set
–  In each voxel, use CT value averaged over
all phases of 4D data set

How can we improve the quality of

4DCT images?
•  Common causes of artifacts on 4DCT
images
–  Inadequate sampling
–  Irregular breathing

gantry rotation time (typically 0.4 sec - 0.5 sec) × resp rate (min -1 )
pitch ≤
60 sec/min

Inadequate sampling

•  Helical: Table translation during one

respiratory cycle must be less than detector
width
gantry rotation time (0.4 s - 0.5 s) × respiratory rate (min -1 )
pitch ≤
60 sec/min

•  Cine: Table must be stationary for at least 1

respiratory cycle + gantry rotation

18
 3/30/12

gantry rotation time (typically 0.4 sec - 0.5 sec) × resp rate (min -1 )
pitch ≤
60 sec/min

Inadequate sampling - helical

•  Early versions of
reconstruction
software displayed
gaps where adequate
information was not
obtained
•  Newer software fills in
gaps – but still may not
be accurate

Inadequate sampling - cine

•  Poor phase match

gantry rotation time (typically 0.4 sec - 0.5 sec) × resp rate (min -1 )
pitch ≤
60 sec/min

Inadequate sampling

•  If appropriate pitch or table motion

cannot be achieved due to slow
respiration, it may be necessary for
patient to increase respiratory rate.

19
 3/30/12

gantry rotation time (typically 0.4 sec - 0.5 sec) × resp rate (min -1 )
pitch ≤
60 sec/min

Irregular breathing

•  4DCT makes the

assumption that patient
respiration is uniform
•  Irregular breathing may
result in artifacts
–  Note irregular anterior
surface of abdomen

Patient training procedure in

preparation of a 4D scan
•  Explain to patient the nature of
breathing procedure – need for
consistent breathing pattern
•  Monitor patient breathing pattern
•  Determine respiratory rate (used to set
scanner pitch) – worst case estimate
•  Determine if respiration is reproducible
enough to acquire a 4DCT

Patient coaching (4 approaches)

1)  Do nothing – many patients breathe best by

being relaxed and not thinking about breathing.
2)  Pre-imaging relaxation and coaching – works
for many nervous patients.
3)  Audio prompting: Works best to modify
breathing rate, but makes nervous patient more
nervous
–  Nonverbal audio device shows promise for these
patients

4)  Video prompting: Helps patients who breathe

irregularly in amplitude.

20
 3/30/12

Patient Video Feedback Devices

•  The patients breathing motion can be fed back

to the patient along with a target amplitude to
help improve the quality of 4DCT imaging.

Post-scan processing

•  Respiratory trace file in RPM system

indicates tags where 0% occurs
•  Tags determined by predictive filter so
there may be some inaccuracy
•  Modify location of tags so that 0%
occurs at true end inspiration

Post-scan processing

•  Displacement binning
–  Bin based on displacement of tumor (or
tumor surrogate) rather than equally-
spaced phases
–  Can be implemented by editing respiratory
trace file before binning
–  Not yet demonstrated to be advantageous
•  Presently, editing of 0% tag is primary
method of post-scan processing

21
 3/30/12

Where do we go from here?

•  Pelc (RSNA 2011) – talk on future

developments in CT
–  Faster gantry speeds – 0.1 sec gantry
rotation
–  Larger detectors – 512 slices (almost there
now)
•  True 4DCT, where 4th dimension is time
•  Removes issues of table translation,
irregular breathing

Take-home message
•  Respiratory-induced tumor motion is
significant and unpredictable
•  Various methods have been developed to
measure the extent of respiratory-induced
tumor motion
•  We are now able to rationally define target
volumes that explicitly account for the effect
of respiratory motion
•  4DCT can become the standard of care for
imaging thoracic tumors.

In 4DCT, what is the 4th dimension?

11% 1.  Elapsed time

0% 2.  Interfractional motion
89% 3.  Phase of the respiratory cycle
0% 4.  Respiratory rate

22
 3/30/12

In 4DCT, what is the 4th dimension?

3.  Phase of the respiratory cycle

Rationale: We combine information at the same phase

of the respiratory cycle from many cycles to
overcome the inability to acquire adequate imaging
data in a single respiratory cycle.

Reference: Starkschall G, Desai N, Balter P, et al.,

Quantitative assessment of four-dimensional
computed tomography image acquisition quality, J
ApplClin Med Phys. 8(3): 1-20 (2007).

Which of the following is NOT involved in

the acquisition of 4DCT images?

0% 1.  Acquiring large amounts of image data during

respiration
13% 2.  Combining image data at the same phase from
several respiratory cycles
3.  Reconstructing image data at specific phases in
6%
the respiratory cycle for each patient location
4.  Using deformable image registration to propagate
81%
the GTV among several respiratory cycles

Which of the following is NOT involved in

the acquisition of 4DCT images?

4.  Using deformable image registration to

propagate the GTV among several
respiratory cycles

Rationale: Deformable image registration is used in planning for

4D dose calculations based on 4D image data sets, and not on
the acquisition of the 4DCT data sets.
Reference: Rietzel E., T. Pan, and G.T.Y. Chen, “Four-dimensional
computed tomography: Image formation and clinical
protocol,” Med Phys 32 (4), 874 – 89 (2005).; Keall P.J. et
al.,The management of respiratory motion in radiation
oncology report of AAPM Task Group 76, Med Phys. 33 (1),
3874 – 900 (2006).

23
 3/30/12

Which of the following is NOT an ICRU-

defined term?

0% 1.  Clinical target volume

2% 2.  Gross tumor volume
91% 3.  Internal gross tumor volume
8% 4.  Internal target volume

Which of the following is NOT an ICRU-

defined term?

3.  Internal gross tumor volume

Rationale: Some institutions have used the concept of

an internal gross tumor volume (IGTV), consisting
of the envelope of motion of the GTV, in the
process of target delineation, but this quantity is
not recognized by the ICRU.
Reference: ICRU Report 50, “Prescribing, Recording,
and Report Photon Beam Therapy”; ICRU Report
62, “Prescribing, Recording and Reporting Photon
Beam Therapy (Supplement to ICRU Report 50)”.

The displacements in the anterior surface of

the abdomen illustrated in this figure resulted
from

70% 1.  Irregular breathing

10% 2.  Too rapid breathing
2% 3.  Too rapid gantry rotation
19% 4.  Too rapid table translation

24
 3/30/12

The displacements in the anterior surface of

the abdomen illustrated in this figure resulted
from
1.  Irregular breathing

Rationale: The gaps in the figure

were the result of too rapid table
translation.
Reference: Starkschall G, Desai N,
Balter P, et al., Quantitative
assessment of four-dimensional
computed tomography image
acquisition quality, J Appl Clin
Med Phys. 8(3): 1-20 (2007).

All of the following methods have been

successfully used to address the problem of
irregular breathing EXCEPT

0% 1.  Audio prompting

67% 2.  Displacement binning
29% 3.  Editing 0% tags
5% 4.  Video feedback

All of the following methods have been

successfully used to address the problem of
irregular breathing EXCEPT
2.  Displacement binning

Rationale: Although, in principle, displacement binning

should assist in improving images obtained during
irregular breathing, in practice this has not yet been
shown to be effective.
Reference: Keall P.J. et al.,The management of
respiratory motion in radiation oncology report of
AAPM Task Group 76, Med Phys. 33 (1), 3874 – 900
(2006).

25
 3/30/12

Thank you

26